Hi,
We couldn’t avoid using unnatural light for taking our photos, so we were wondering how that will work for when we are generating the cone catch model. In the video it is done by taking a picture of the colour standard under natural light. For our situation to generate cone catch model for the camera should we take a photo of the colour standard under natural light or the light the photos were taken under? We know that the spectrum of the lights we are using differs from natural light spectrum. For generating our model we are using the colour checker standard and the values found in the mica tool box for the colour checker reflectance (each standard having its value found by averaging its visible light values), is it okay to use those values of reflectance under unnatural light?
Thank you for your time and help!
Sadly it’s not straightforward moving between light sources using a colour chart, even if you know the spectral emissions of your lights. One source of error is fluorescence in the chart. Even high-quality colour charts have fluorescence in some of the colours, so the simple laws of reflectance which many of our models depend on start to fall apart.
If you’re using a light source which does a good job of recreating natural sunlight then I would guess that this source of error is very low, but I simply haven’t done the testing or modelling required to say how much of a difference this would cause.
However, if you’ve already got the images then you could do this:
- Carry on analysing the data using D65 cone catch models even though the photos were taken under artificial light
- Model the colour differences of the artificial light by photographing the colour chart under natural light and your artificial light.
- Convert both these images to cone-catch using the same model (using standard mspec image calibration)
- Measure the chart colours in each of these images
- Plot the colours against each-other. e.g. LW (natural light) vs LW (artificial light). Calculate the R^2 value for this correlation and all other receptors. Hopefully the two light sources result in cone-catch values which are very similar!
- Report the above steps in any publication, including the R^2 values.
Cheers,
Jolyon